Manajemen | Fakultas Ekonomi Universitas Maritim Raja Ali Haji 450.full
Information, Avoidance Behavior, and
Health
The Effect of Ozone on Asthma
Hospitalizations
Matthew Neidell
abstract
This paper assesses whether responses to information about risk impact
estimates of the relationship between ozone and asthma in Southern
California. Using a regression discontinuity design, I find smog alerts
significantly reduce daily attendance at two major outdoor facilities. Using
daily time-series regression models that include year-month and small area
fixed effects, I find estimates of the effect of ozone for children and the elderly
that include information are significantly larger than estimates that do not.
These results are consistent with the hypothesis that individuals take
substantial action to reduce exposure to risk; estimates ignoring these actions
are severely biased.
I. Introduction
Information about risk is provided in numerous areas of public health
so individuals can adjust their behavior when confronted with risk. Observational
analyses that estimate health effects from exposure to these risks, however, typically
assume no behavioral responses to risk. For example, several recent studies
Matthew Neidell is an assistant professor of health policy and management at Columbia University. The
author thanks Janet Currie, Michael Greenstone, Ken Chay, Sherry Glied, Enrico Moretti, Helen Levy,
Will Manning, Tomas Philipson, Sylvia Brandt, Elizabeth Powers, Michael Khoo, Joshua Graff Zivin,
Amanda Lang, Paul Rathouz, Bob Kaestner, Kerry Smith, Pat Bayer, Wes Hartmann, an anonymous
referee, and numerous seminar participants for valuable feedback. He is also very grateful to Bruce
Selik and Joe Cassmassi of the South Coast Air Quality Management District for providing information
on smog alerts, Mei Kwan, E. C. Krupp, and Ken Warren for help with obtaining the attendance data
used for this analysis, and Sarah Kishinevsky, Mike Kraft, and Sonalini Singh for excellent research
assistance. Financial support from the University of Chicago’s Center for Integrating Statistical and
Environmental Science is gratefully acknowledged. Some data used in this article are available from
October 2009 through September 2012, while information on obtaining the confidential data will be
provided upon request from Matthew Neidell, Columbia University, 600 W. 168th Street, 6th floor, New
York, NY 10032 .
½Submitted April 2007; accepted September 2007
T H E JO U R NAL O F H U M A N R ES O U R C ES
d
44
d
2
Neidell
examining the effect of outdoor air pollution on health provide convincing empirical
evidence of exogenous variation in ambient pollution, but do not account for actions
individuals may take to reduce their exposure to pollution, such as decreasing
the amount of time spent outside (for example, Friedman et al. 2001; Chay and
Greenstone 2003a,b; Currie and Neidell 2005). Even if ambient pollution levels
are as good as randomly assigned, exposure to pollution remains endogenous, so
estimates of the biological effect of pollution on health are biased.
This paper assesses whether behavioral responses to information about risk impact
estimates of the relationship between ozone and asthma hospitalizations in Southern California. Avoidance behavior is especially pervasive in this context because information
about pollution is so widespread. The Los Angeles Times, the most widely circulated
newspaper in the region, provides the pollutant standards index (PSI),1 a continuous index that converts forecasted daily pollution levels into an easily understandable format
and, depending on the level of pollution, advises the public regarding health effects
and precautionary actions to take (Environmental Protection Agency 1999). Furthermore, various media channels, including television and radio, announce ‘‘smog alerts’’
when forecasted ambient ozone exceeds a particular threshold. These alerts encourage
susceptible individuals, such as children and the elderly, to avoid exposure to ozone by
remaining indoors and all others to avoid rigorous outdoor activity. Importantly, both
sources of information are forecasted a day in advance to provide ample time for individuals to react.
To assess the impact of responses to information about risk, this paper proceeds in
two stages. First, I estimate whether people respond to information about pollution
by examining the effect of smog alerts on daily outdoor activities. As a measure of outdoor activities, I use daily attendance from 1989–97 at two major outdoor facilities in
Southern California: the Los Angeles Zoo and Griffith Park Observatory. To identify
the effect of smog alerts, I employ a regression discontinuity design by exploiting
the deterministic selection rule used for issuing alerts. If days just above or below this
threshold do not vary systematically with outdoor activity decisions, then I can obtain
estimates of the causal effect of alerts on avoidance behavior. In support of this approach, other observable characteristics, such as weather and observed pollution, move
smoothly around this threshold, suggesting any change in outdoor activities at this
threshold can be directly attributed to smog alerts.
The second stage of this paper focuses on estimating the bias from incorporating
these responses by examining the impact of ozone on asthma hospitalizations. If individuals respond to information about ozone and ozone affects health, then accounting
for responses should increase the estimated effect of ozone. To identify the effect of
ozone, I estimate daily time-series regression models that include year-month fixed
effects and finely defined geographic area fixed effects. The year-month fixed effects
nonparametrically absorb seasonal and temporal trends in ozone and health. The area
fixed effects capture observed and unobserved factors common to residents within that
area, such as income, education, and access to health insurance, to the extent they do not
vary over time. I also include extensive controls for weather, other outdoor pollutants,
and day of week dummy variables to capture time varying factors. The remaining
1. The Air Quality Index replaced the PSI in 1999, but data in this analysis goes through 1997.
451
452
The Journal of Human Resources
variation in ozone is likely independent from the numerous behavioral and environmental factors that affect health.
The first set of results indicates that people respond to smog alerts. Attendance is significantly lower on days when smog alerts are announced, with declines in preferred
specifications of 6 and 13 percent across the two facilities considered. These results
are generally insensitive to functional form assumptions of the regression discontinuity
design and are robust to several specification checks. For example, the estimates are remarkably unaffected by the inclusion of numerous controls for weather conditions and
observed air quality, both significant predictors of outdoor activities. Attendance for
children and the elderly, two groups specifically targeted by air quality information, display greater responses to alerts. These findings indicate that people value the provision
of information contained in the warnings.
The second set of results confirms that accounting for responses to information about
risk drastically alters conclusions about the relationship between ozone and health,
though only for susceptible populations. Ignoring avoidance behavior suggests ozone
has the smallest effect on children, while controlling for avoidance behavior indicates
the biggest effect on children. Including smog alerts and the PSI increases estimates of
the effect of ozone by roughly 160 percent for children and 40 percent for the elderly,
but has no effect on estimates for adults. These patterns by age are consistent with children and the elderly having greater response to alerts than adults.
The coefficients on information about risk can also be directly interpreted for understanding the effect of ozone on health. If providing information improves health ceteris
paribus—importantly, holding observed ozone levels fixed—this implies evidence of
both a response to information and an effect of ozone on health. For example, if individuals respond to an alert by reducing their exposure to ozone and, consequently, their
health improves, then ozone must affect health. I find that information has a significant,
negative effect on hospitalizations for children and the elderly, further supporting that
ozone effects health.
Evidence that people respond to public information about risk is interesting in own
right. First, studies consistently document that individuals adjust their behavior in response to information about permanent, long-standing risk.2 Information is often disseminated to shape individuals’ short-run behavior in response to imminent dangers,
such as terrorism threats, environmental hazards, and disease outbreaks, and these
findings indicate that people respond to rapidly changing information about risk. Second, economic epidemiology models imply that diseases can be self-limiting through
avoidance behavior, so there may be little role for government intervention (Philipson 2000). Studies that examine behavioral responses to disease outbreaks, however, are unable to distinguish how information about risk is transmitted as a
disease spreads, whether by private or public information.3 Because air quality information is publicly provided, these findings indicate the government has a potential
role in affecting illness by disseminating information about risk.
Most importantly, these results are consistent with the hypothesis that individuals
take substantial actions to reduce exposure to ozone; estimates of the health effect of
ozone that ignore these actions are severely biased. Given the contentious debates
2. For example, see Ippolito and Mathios (1990), Jin and Leslie (2003), and Smith and Johnson (1988).
3. For example, see Ahituv, Hotz, and Philipson (1996), Mullahy (1999), and Philipson (1996).
Neidell
surrounding air quality regulations,4 it is important to account for this behavior to correctly measure the costs of ozone to society. Moreover, given the prevalent availability
of information about risk, these findings suggests the importance of accounting for protective behavior in a wide range of statistical analyses that rely on observational data to
understand biological relationships.
II. Background
A. Air Quality and Health
One of the most influential studies on air pollution and health comes from the London
‘‘killer’’ fog, a severe air pollution episode in London, England, in December, 1952
(Wilson and Spengler 1996). The unexpected—and arguably exogenous—change in
air quality stemming from a temperature inversion has been linked with thousands of
premature deaths (Logan and Glasg 1953). Since air quality was not regularly monitored and its health effects were poorly understood (and fog was common in London),
people were largely uninformed of the harm they faced from exposure, so there was limited, if any, avoidance behavior.
More recent studies have focused on determining effects from air pollution at the
lower levels commonly found in developed countries. Countless observational analyses
have estimated statistical associations between pollution and various health outcomes,
mostly focusing on mortality and respiratory related morbidity, such as asthma, with
summaries provided in Levy et al. (2001), Brunekreef and Holgate (2002), Wilson
and Spengler (1996), and Environmental Protection Agency (2006). A common statistical approach is the use of daily time-series data on health outcomes, such as hospitalizations, linked with contemporaneous and lagged levels of pollution and potential
confounding variables, such as weather (for example, Katsouyanni et al. 1996, Thurston et al. 1992, Schwartz 1995, Bell et al. 2004). An alternative approach is cohort
studies that follow individuals over time and compare pollution measures aggregated
over time with health outcomes (for example, Dockery et al. 1993, Pope et al. 1995,
Kinney, Thurston, and Raizenne 1996).
These approaches have been criticized on the grounds that ambient pollution is not
randomly assigned (Chay and Greenstone 2003a,b), leading to a surge in quasi-experimental techniques to isolate exogenous changes in pollution. For example, several
studies by Pope and others (Pope 1989, Pope, Schwartz, and Ransom 1992, Ransom
and Pope 1995) used changes in pollution that resulted from the opening and closing
of a steel mill in Utah because of a labor strike. Heinrich, Hoelscher, and Wichmann
(2000) exploited changes in total suspended particles (TSPs) in Eastern Germany from
shifts in industrial activity and stricter air quality regulations as a result of German
reunification in 1990. Friedman et al. (2001) used the temporary change in air quality
from short-term traffic rules intended to accommodate the 1996 Olympic Games in
Atlanta, GA. Two studies by Chay and Greenstone (2003a,b) exploited the implementation of the Clean Air Act of 1970 and the recession of the early 1980s that induced
4. For example, new air quality standards for ozone were proposed in 1997 and upheld in court until 2003.
453
454
The Journal of Human Resources
considerable temporal and spatial variation in TSPs throughout the United States.
Lederman et al. (2004) compared birth outcomes of children born to women living
or working within two miles of the World Trade Center who were pregnant on September 11, 2001, to children born to women more than two miles away. Lleras-Muney
(2005) used the plausibly random relocation of military personnel to new residential
locations to estimate the effect of various pollutants on children’s health.5
Although these studies often go to great lengths to provide convincing empirical
evidence that ambient pollution levels can be viewed as randomly assigned, they typically assume people do not respond to their assigned level of pollution. Unlike the
early killer fog episodes where people were largely unaware of the danger associated
with pollution levels, the scope for avoidance behavior has increased considerably
over time, though the extent to which it exists may vary depending on the specific
context. If people respond to higher pollution levels by increasing avoidance behavior, then the estimated effect of pollution on health is biased down.
B. Ozone, Asthma, and Information
Ground-level ozone is a criteria pollutant6 regulated under the Clean Air Acts that
affects respiratory morbidity by irritating lung airways, decreasing lung function,
and increasing respiratory symptoms, with effects exacerbated for susceptible individuals, such as children, the elderly, and particularly asthmatics. Because it takes some
time for a disease such as asthma to progress, symptoms can arise from contemporaneous exposure (in as quickly as one hour of exposure), from cumulative exposure over
several days, or several days after exposure. For example, ‘‘an asthmatic may be impacted by ozone on the first day of exposure, have further effects triggered on the second
day, and then report to the emergency room for an asthmatic attack three days after exposure’’ (Environmental Protection Agency 2006).
Numerous studies have documented associations between ozone and hospital
admissions for asthma using daily time series data—the approach taken in this
paper—with the most comprehensive review provided in Environmental Protection
Agency (2006). Evidence generally supports significant increases in asthma hospitalizations as ozone increases, though it is not uncommon to find no association (for
example, Norris et al. 1999; Lierl and Hornung 2003; Garty et al. 1998). Cumulative
effects are supported by the data, with the largest effect on hospital admissions from
one lag of ozone, followed by two lags, then zero, three, and four lags in a virtual tie,
with no effect established beyond four lags (Environmental Protection Agency
2006).
The process leading to ozone formation makes it highly predictable and straightforward to avoid. Ozone is not directly emitted into the atmosphere, but is formed
from interactions of nitrogen oxides (NOx) and volatile organic compounds (VOCs)
(both of which are directly emitted) in the presence of heat, sunlight, and solar radiation. Because of this process, ozone levels vary considerably both across and within
days, peaking in the summer and middle of the day when heat, sunlight, and solar
5. For examples of other studies, see Currie and Neidell (2005), Neidell (2004), Jayachandran (2005), and
Sneeringer (2006).
6. Criteria pollutants are six common air pollutants with established health-based air quality standards.
They include ozone, carbon monoxide, particulate matter, nitrogen dioxide, lead, and sulfur dioxide.
Neidell
radiation are highest (Environmental Protection Agency 2006). Therefore, ozone levels can be predicted using weather forecasts, and ozone rapidly breaks down indoors
because there is less heat and sunlight (Chang et al. 2000). Since symptoms from
ozone exposure can arise over a short period of time, altering short-run exposure
by going indoors can reduce the onset of symptoms.
Because of the potential effectiveness of avoidance behavior, the pollutant standards index (PSI) was developed to inform the public of expected air quality conditions. The PSI, which is forecasted on a daily basis, ranges from 0–500 and is
indexed so that a value of 100 corresponds to the National Ambient Air Quality
Standards set forth in the Clean Air Acts. The PSI is computed for five of the criteria
pollutants, and the maximum PSI across pollutants is required by federal law to be
reported in major newspapers (Environmental Protection Agency 1999). It also contains a brief legend summarizing the air quality: 0–50 good; 51–100 moderate; 101–
200 unhealthful; 201–275 very unhealthful; and 275+ hazardous.
In addition to providing the PSI, California state law requires the announcement of
a stage I air quality episode when the PSI is at least 200, which corresponds to 0.20
parts per million (ppm) for ozone.7 These episodes, which also occur on a daily basis, are more widely publicized than the PSI; they are announced on both television
and radio. Although air quality episodes can potentially be issued for any criteria pollutant, they have only been issued for ozone. Compatible with seasonal patterns of
ozone, alerts are issued from March 1 to October 31. Because ozone is a major component of urban smog, this has given rise to the term ‘‘smog alerts.’’
The agency that provides air quality forecasts and issues smog alerts for Southern
California is the South Coast Air Quality Management District (SCAQMD). They
produce the following day’s air quality forecast by noon the day before to provide
enough time to disseminate the information. Because SCAQMD covers all of Orange
County and the most populated parts of Los Angeles, Riverside, and San Bernardino
counties—an area with considerable spatial variation in ozone—they provide the
forecast for each of the 38 source receptor areas (SRAs) within SCAQMD. When
an alert is issued, the staff at SCAQMD contacts a set list of recipients, including
local schools and newspapers. The media then further circulate the information to
the public, but greatly condense it. For example, the Los Angeles Times, although
it receives air quality forecasts for all 38 SRAs in SCAQMD, only reports air quality
forecasts for ten air monitoring areas (AMAs) in SCAQMD by taking the maximum
forecasted value of the SRAs within an AMA.
Given the reporting process and the factors believed to affect ozone formation, the
model used for issuing an alert can be summarized as:
ð1Þ alertat ¼ 1fmaxat ðozone fst ¼ f ðweather fst; ozonest21; solradt ÞÞ $ 0:20g
where the subscripts a, s, and t indicate AMA, SRA, and date, respectively, ozonef is
the forecasted one-hour level of ozone, weatherf is the weather forecast, ozone is observed one-hour ozone, solrad is solar radiation, and 1fg is an indicator function
equal to one when the forecasted ozone exceeds 0.20 ppm and 0 otherwise.
7. Additionally, a Stage II air quality episode is issued when the PSI exceeds 250 or ozone forecast exceeds
0.30 ppm, but this only occurred once over the time period studied.
455
456
The Journal of Human Resources
III. Data
A. Outdoor attendance
For a measure of time spent outdoors, accurately recorded individual level time diaries would be an ideal source of data, but such data are generally unavailable at a
daily level over a sufficient period of time. I instead use daily aggregate measures of
attendance at two outdoor facilities within the boundaries of the SCAQMD: the Los
Angeles Zoo and Botanical Gardens (‘‘Zoo’’) and Griffith Park Observatory (‘‘Observatory’’).8 The Zoo is open daily from 10 a.m. to 5 p.m., with the closing time
extended to 6 p.m. from July 1 to Labor Day. The Observatory is open from 2
p.m. to 10 p.m. Tuesday through Friday and 12:30 p.m. to 10 p.m. on Saturday
and Sunday, but is open from 12:30 p.m. to 10 p.m. everyday during the summer.
Both are located a short distance from downtown Los Angeles and attract sizeable
crowds, averaging more than 4,500 attendees per day. These data are available from
1989–97, with descriptive statistics for each shown in Table 1.
Although focusing on two specific facilities limits generalizability, these data provide several advantages over time use surveys. One, because they are administrative
data, they are likely to be free of recall errors that often arise in survey data. Two,
these data are available over a long period of time in which there is substantial variation in ozone levels, forecasts, and smog alerts. Three, the exact dates are available
in the attendance data, allowing me to merge several files at the daily levels and use
the regression discontinuity design. Therefore, this approach improves upon measurement, precision, and causality at the expense of generalizability.
The Zoo charges varying admission fees, so it also offers some breakdown of attendance by demographics to assess heterogeneity in responses to alerts. Separate
daily attendance is available for adults, children younger than 2, children aged
2–12, and seniors aged 62 and older, with means for each presented in Table 1. Children and the elderly are two groups considered susceptible to the effects of ozone, so
their responses may differ from adults.9 Although this definition of susceptibility is
not exhaustive, both groups are specifically targeted by air quality information. The
Zoo also offers attendance for Greater Los Angeles Zoo Association (GLAZA) members. While the Zoo is both a tourist and local attraction, GLAZA members are typically local residents who may be more aware of alerts and find it easier to switch
activities.10 Therefore, they are more likely to respond to alerts, providing one robustness check of the model.
I also collect data on attendance at baseball games (available in the ‘‘game logs’’
at www.retrosheet.org) for two major league baseball teams in SCAQMD: the Los
Angeles Dodgers, who play a short distance form downtown Los Angeles, and the
California Angels,11 who play in Anaheim, approximately 30 miles southeast of
8. The Zoo and Observatory are owned and operated by the City of Los Angeles.
9. Children’s responses to alerts could reflect state rules requiring schools to reschedule outdoor activities,
but the analysis focuses almost exclusively on summer months when children are not in school and also
includes a dummy variable for summer schedule.
10. GLAZA members pay an annual fee and do not pay an admission fee per visit.
11. The California Angels were renamed the Anaheim Angels in 1997, and then the Los Angeles Angels of
Anaheim in 2005.
Neidell
Table 1
Summary statistics
A. Daily attendance
Zoo (n ¼ 1949)
Children
Health
The Effect of Ozone on Asthma
Hospitalizations
Matthew Neidell
abstract
This paper assesses whether responses to information about risk impact
estimates of the relationship between ozone and asthma in Southern
California. Using a regression discontinuity design, I find smog alerts
significantly reduce daily attendance at two major outdoor facilities. Using
daily time-series regression models that include year-month and small area
fixed effects, I find estimates of the effect of ozone for children and the elderly
that include information are significantly larger than estimates that do not.
These results are consistent with the hypothesis that individuals take
substantial action to reduce exposure to risk; estimates ignoring these actions
are severely biased.
I. Introduction
Information about risk is provided in numerous areas of public health
so individuals can adjust their behavior when confronted with risk. Observational
analyses that estimate health effects from exposure to these risks, however, typically
assume no behavioral responses to risk. For example, several recent studies
Matthew Neidell is an assistant professor of health policy and management at Columbia University. The
author thanks Janet Currie, Michael Greenstone, Ken Chay, Sherry Glied, Enrico Moretti, Helen Levy,
Will Manning, Tomas Philipson, Sylvia Brandt, Elizabeth Powers, Michael Khoo, Joshua Graff Zivin,
Amanda Lang, Paul Rathouz, Bob Kaestner, Kerry Smith, Pat Bayer, Wes Hartmann, an anonymous
referee, and numerous seminar participants for valuable feedback. He is also very grateful to Bruce
Selik and Joe Cassmassi of the South Coast Air Quality Management District for providing information
on smog alerts, Mei Kwan, E. C. Krupp, and Ken Warren for help with obtaining the attendance data
used for this analysis, and Sarah Kishinevsky, Mike Kraft, and Sonalini Singh for excellent research
assistance. Financial support from the University of Chicago’s Center for Integrating Statistical and
Environmental Science is gratefully acknowledged. Some data used in this article are available from
October 2009 through September 2012, while information on obtaining the confidential data will be
provided upon request from Matthew Neidell, Columbia University, 600 W. 168th Street, 6th floor, New
York, NY 10032 .
½Submitted April 2007; accepted September 2007
T H E JO U R NAL O F H U M A N R ES O U R C ES
d
44
d
2
Neidell
examining the effect of outdoor air pollution on health provide convincing empirical
evidence of exogenous variation in ambient pollution, but do not account for actions
individuals may take to reduce their exposure to pollution, such as decreasing
the amount of time spent outside (for example, Friedman et al. 2001; Chay and
Greenstone 2003a,b; Currie and Neidell 2005). Even if ambient pollution levels
are as good as randomly assigned, exposure to pollution remains endogenous, so
estimates of the biological effect of pollution on health are biased.
This paper assesses whether behavioral responses to information about risk impact
estimates of the relationship between ozone and asthma hospitalizations in Southern California. Avoidance behavior is especially pervasive in this context because information
about pollution is so widespread. The Los Angeles Times, the most widely circulated
newspaper in the region, provides the pollutant standards index (PSI),1 a continuous index that converts forecasted daily pollution levels into an easily understandable format
and, depending on the level of pollution, advises the public regarding health effects
and precautionary actions to take (Environmental Protection Agency 1999). Furthermore, various media channels, including television and radio, announce ‘‘smog alerts’’
when forecasted ambient ozone exceeds a particular threshold. These alerts encourage
susceptible individuals, such as children and the elderly, to avoid exposure to ozone by
remaining indoors and all others to avoid rigorous outdoor activity. Importantly, both
sources of information are forecasted a day in advance to provide ample time for individuals to react.
To assess the impact of responses to information about risk, this paper proceeds in
two stages. First, I estimate whether people respond to information about pollution
by examining the effect of smog alerts on daily outdoor activities. As a measure of outdoor activities, I use daily attendance from 1989–97 at two major outdoor facilities in
Southern California: the Los Angeles Zoo and Griffith Park Observatory. To identify
the effect of smog alerts, I employ a regression discontinuity design by exploiting
the deterministic selection rule used for issuing alerts. If days just above or below this
threshold do not vary systematically with outdoor activity decisions, then I can obtain
estimates of the causal effect of alerts on avoidance behavior. In support of this approach, other observable characteristics, such as weather and observed pollution, move
smoothly around this threshold, suggesting any change in outdoor activities at this
threshold can be directly attributed to smog alerts.
The second stage of this paper focuses on estimating the bias from incorporating
these responses by examining the impact of ozone on asthma hospitalizations. If individuals respond to information about ozone and ozone affects health, then accounting
for responses should increase the estimated effect of ozone. To identify the effect of
ozone, I estimate daily time-series regression models that include year-month fixed
effects and finely defined geographic area fixed effects. The year-month fixed effects
nonparametrically absorb seasonal and temporal trends in ozone and health. The area
fixed effects capture observed and unobserved factors common to residents within that
area, such as income, education, and access to health insurance, to the extent they do not
vary over time. I also include extensive controls for weather, other outdoor pollutants,
and day of week dummy variables to capture time varying factors. The remaining
1. The Air Quality Index replaced the PSI in 1999, but data in this analysis goes through 1997.
451
452
The Journal of Human Resources
variation in ozone is likely independent from the numerous behavioral and environmental factors that affect health.
The first set of results indicates that people respond to smog alerts. Attendance is significantly lower on days when smog alerts are announced, with declines in preferred
specifications of 6 and 13 percent across the two facilities considered. These results
are generally insensitive to functional form assumptions of the regression discontinuity
design and are robust to several specification checks. For example, the estimates are remarkably unaffected by the inclusion of numerous controls for weather conditions and
observed air quality, both significant predictors of outdoor activities. Attendance for
children and the elderly, two groups specifically targeted by air quality information, display greater responses to alerts. These findings indicate that people value the provision
of information contained in the warnings.
The second set of results confirms that accounting for responses to information about
risk drastically alters conclusions about the relationship between ozone and health,
though only for susceptible populations. Ignoring avoidance behavior suggests ozone
has the smallest effect on children, while controlling for avoidance behavior indicates
the biggest effect on children. Including smog alerts and the PSI increases estimates of
the effect of ozone by roughly 160 percent for children and 40 percent for the elderly,
but has no effect on estimates for adults. These patterns by age are consistent with children and the elderly having greater response to alerts than adults.
The coefficients on information about risk can also be directly interpreted for understanding the effect of ozone on health. If providing information improves health ceteris
paribus—importantly, holding observed ozone levels fixed—this implies evidence of
both a response to information and an effect of ozone on health. For example, if individuals respond to an alert by reducing their exposure to ozone and, consequently, their
health improves, then ozone must affect health. I find that information has a significant,
negative effect on hospitalizations for children and the elderly, further supporting that
ozone effects health.
Evidence that people respond to public information about risk is interesting in own
right. First, studies consistently document that individuals adjust their behavior in response to information about permanent, long-standing risk.2 Information is often disseminated to shape individuals’ short-run behavior in response to imminent dangers,
such as terrorism threats, environmental hazards, and disease outbreaks, and these
findings indicate that people respond to rapidly changing information about risk. Second, economic epidemiology models imply that diseases can be self-limiting through
avoidance behavior, so there may be little role for government intervention (Philipson 2000). Studies that examine behavioral responses to disease outbreaks, however, are unable to distinguish how information about risk is transmitted as a
disease spreads, whether by private or public information.3 Because air quality information is publicly provided, these findings indicate the government has a potential
role in affecting illness by disseminating information about risk.
Most importantly, these results are consistent with the hypothesis that individuals
take substantial actions to reduce exposure to ozone; estimates of the health effect of
ozone that ignore these actions are severely biased. Given the contentious debates
2. For example, see Ippolito and Mathios (1990), Jin and Leslie (2003), and Smith and Johnson (1988).
3. For example, see Ahituv, Hotz, and Philipson (1996), Mullahy (1999), and Philipson (1996).
Neidell
surrounding air quality regulations,4 it is important to account for this behavior to correctly measure the costs of ozone to society. Moreover, given the prevalent availability
of information about risk, these findings suggests the importance of accounting for protective behavior in a wide range of statistical analyses that rely on observational data to
understand biological relationships.
II. Background
A. Air Quality and Health
One of the most influential studies on air pollution and health comes from the London
‘‘killer’’ fog, a severe air pollution episode in London, England, in December, 1952
(Wilson and Spengler 1996). The unexpected—and arguably exogenous—change in
air quality stemming from a temperature inversion has been linked with thousands of
premature deaths (Logan and Glasg 1953). Since air quality was not regularly monitored and its health effects were poorly understood (and fog was common in London),
people were largely uninformed of the harm they faced from exposure, so there was limited, if any, avoidance behavior.
More recent studies have focused on determining effects from air pollution at the
lower levels commonly found in developed countries. Countless observational analyses
have estimated statistical associations between pollution and various health outcomes,
mostly focusing on mortality and respiratory related morbidity, such as asthma, with
summaries provided in Levy et al. (2001), Brunekreef and Holgate (2002), Wilson
and Spengler (1996), and Environmental Protection Agency (2006). A common statistical approach is the use of daily time-series data on health outcomes, such as hospitalizations, linked with contemporaneous and lagged levels of pollution and potential
confounding variables, such as weather (for example, Katsouyanni et al. 1996, Thurston et al. 1992, Schwartz 1995, Bell et al. 2004). An alternative approach is cohort
studies that follow individuals over time and compare pollution measures aggregated
over time with health outcomes (for example, Dockery et al. 1993, Pope et al. 1995,
Kinney, Thurston, and Raizenne 1996).
These approaches have been criticized on the grounds that ambient pollution is not
randomly assigned (Chay and Greenstone 2003a,b), leading to a surge in quasi-experimental techniques to isolate exogenous changes in pollution. For example, several
studies by Pope and others (Pope 1989, Pope, Schwartz, and Ransom 1992, Ransom
and Pope 1995) used changes in pollution that resulted from the opening and closing
of a steel mill in Utah because of a labor strike. Heinrich, Hoelscher, and Wichmann
(2000) exploited changes in total suspended particles (TSPs) in Eastern Germany from
shifts in industrial activity and stricter air quality regulations as a result of German
reunification in 1990. Friedman et al. (2001) used the temporary change in air quality
from short-term traffic rules intended to accommodate the 1996 Olympic Games in
Atlanta, GA. Two studies by Chay and Greenstone (2003a,b) exploited the implementation of the Clean Air Act of 1970 and the recession of the early 1980s that induced
4. For example, new air quality standards for ozone were proposed in 1997 and upheld in court until 2003.
453
454
The Journal of Human Resources
considerable temporal and spatial variation in TSPs throughout the United States.
Lederman et al. (2004) compared birth outcomes of children born to women living
or working within two miles of the World Trade Center who were pregnant on September 11, 2001, to children born to women more than two miles away. Lleras-Muney
(2005) used the plausibly random relocation of military personnel to new residential
locations to estimate the effect of various pollutants on children’s health.5
Although these studies often go to great lengths to provide convincing empirical
evidence that ambient pollution levels can be viewed as randomly assigned, they typically assume people do not respond to their assigned level of pollution. Unlike the
early killer fog episodes where people were largely unaware of the danger associated
with pollution levels, the scope for avoidance behavior has increased considerably
over time, though the extent to which it exists may vary depending on the specific
context. If people respond to higher pollution levels by increasing avoidance behavior, then the estimated effect of pollution on health is biased down.
B. Ozone, Asthma, and Information
Ground-level ozone is a criteria pollutant6 regulated under the Clean Air Acts that
affects respiratory morbidity by irritating lung airways, decreasing lung function,
and increasing respiratory symptoms, with effects exacerbated for susceptible individuals, such as children, the elderly, and particularly asthmatics. Because it takes some
time for a disease such as asthma to progress, symptoms can arise from contemporaneous exposure (in as quickly as one hour of exposure), from cumulative exposure over
several days, or several days after exposure. For example, ‘‘an asthmatic may be impacted by ozone on the first day of exposure, have further effects triggered on the second
day, and then report to the emergency room for an asthmatic attack three days after exposure’’ (Environmental Protection Agency 2006).
Numerous studies have documented associations between ozone and hospital
admissions for asthma using daily time series data—the approach taken in this
paper—with the most comprehensive review provided in Environmental Protection
Agency (2006). Evidence generally supports significant increases in asthma hospitalizations as ozone increases, though it is not uncommon to find no association (for
example, Norris et al. 1999; Lierl and Hornung 2003; Garty et al. 1998). Cumulative
effects are supported by the data, with the largest effect on hospital admissions from
one lag of ozone, followed by two lags, then zero, three, and four lags in a virtual tie,
with no effect established beyond four lags (Environmental Protection Agency
2006).
The process leading to ozone formation makes it highly predictable and straightforward to avoid. Ozone is not directly emitted into the atmosphere, but is formed
from interactions of nitrogen oxides (NOx) and volatile organic compounds (VOCs)
(both of which are directly emitted) in the presence of heat, sunlight, and solar radiation. Because of this process, ozone levels vary considerably both across and within
days, peaking in the summer and middle of the day when heat, sunlight, and solar
5. For examples of other studies, see Currie and Neidell (2005), Neidell (2004), Jayachandran (2005), and
Sneeringer (2006).
6. Criteria pollutants are six common air pollutants with established health-based air quality standards.
They include ozone, carbon monoxide, particulate matter, nitrogen dioxide, lead, and sulfur dioxide.
Neidell
radiation are highest (Environmental Protection Agency 2006). Therefore, ozone levels can be predicted using weather forecasts, and ozone rapidly breaks down indoors
because there is less heat and sunlight (Chang et al. 2000). Since symptoms from
ozone exposure can arise over a short period of time, altering short-run exposure
by going indoors can reduce the onset of symptoms.
Because of the potential effectiveness of avoidance behavior, the pollutant standards index (PSI) was developed to inform the public of expected air quality conditions. The PSI, which is forecasted on a daily basis, ranges from 0–500 and is
indexed so that a value of 100 corresponds to the National Ambient Air Quality
Standards set forth in the Clean Air Acts. The PSI is computed for five of the criteria
pollutants, and the maximum PSI across pollutants is required by federal law to be
reported in major newspapers (Environmental Protection Agency 1999). It also contains a brief legend summarizing the air quality: 0–50 good; 51–100 moderate; 101–
200 unhealthful; 201–275 very unhealthful; and 275+ hazardous.
In addition to providing the PSI, California state law requires the announcement of
a stage I air quality episode when the PSI is at least 200, which corresponds to 0.20
parts per million (ppm) for ozone.7 These episodes, which also occur on a daily basis, are more widely publicized than the PSI; they are announced on both television
and radio. Although air quality episodes can potentially be issued for any criteria pollutant, they have only been issued for ozone. Compatible with seasonal patterns of
ozone, alerts are issued from March 1 to October 31. Because ozone is a major component of urban smog, this has given rise to the term ‘‘smog alerts.’’
The agency that provides air quality forecasts and issues smog alerts for Southern
California is the South Coast Air Quality Management District (SCAQMD). They
produce the following day’s air quality forecast by noon the day before to provide
enough time to disseminate the information. Because SCAQMD covers all of Orange
County and the most populated parts of Los Angeles, Riverside, and San Bernardino
counties—an area with considerable spatial variation in ozone—they provide the
forecast for each of the 38 source receptor areas (SRAs) within SCAQMD. When
an alert is issued, the staff at SCAQMD contacts a set list of recipients, including
local schools and newspapers. The media then further circulate the information to
the public, but greatly condense it. For example, the Los Angeles Times, although
it receives air quality forecasts for all 38 SRAs in SCAQMD, only reports air quality
forecasts for ten air monitoring areas (AMAs) in SCAQMD by taking the maximum
forecasted value of the SRAs within an AMA.
Given the reporting process and the factors believed to affect ozone formation, the
model used for issuing an alert can be summarized as:
ð1Þ alertat ¼ 1fmaxat ðozone fst ¼ f ðweather fst; ozonest21; solradt ÞÞ $ 0:20g
where the subscripts a, s, and t indicate AMA, SRA, and date, respectively, ozonef is
the forecasted one-hour level of ozone, weatherf is the weather forecast, ozone is observed one-hour ozone, solrad is solar radiation, and 1fg is an indicator function
equal to one when the forecasted ozone exceeds 0.20 ppm and 0 otherwise.
7. Additionally, a Stage II air quality episode is issued when the PSI exceeds 250 or ozone forecast exceeds
0.30 ppm, but this only occurred once over the time period studied.
455
456
The Journal of Human Resources
III. Data
A. Outdoor attendance
For a measure of time spent outdoors, accurately recorded individual level time diaries would be an ideal source of data, but such data are generally unavailable at a
daily level over a sufficient period of time. I instead use daily aggregate measures of
attendance at two outdoor facilities within the boundaries of the SCAQMD: the Los
Angeles Zoo and Botanical Gardens (‘‘Zoo’’) and Griffith Park Observatory (‘‘Observatory’’).8 The Zoo is open daily from 10 a.m. to 5 p.m., with the closing time
extended to 6 p.m. from July 1 to Labor Day. The Observatory is open from 2
p.m. to 10 p.m. Tuesday through Friday and 12:30 p.m. to 10 p.m. on Saturday
and Sunday, but is open from 12:30 p.m. to 10 p.m. everyday during the summer.
Both are located a short distance from downtown Los Angeles and attract sizeable
crowds, averaging more than 4,500 attendees per day. These data are available from
1989–97, with descriptive statistics for each shown in Table 1.
Although focusing on two specific facilities limits generalizability, these data provide several advantages over time use surveys. One, because they are administrative
data, they are likely to be free of recall errors that often arise in survey data. Two,
these data are available over a long period of time in which there is substantial variation in ozone levels, forecasts, and smog alerts. Three, the exact dates are available
in the attendance data, allowing me to merge several files at the daily levels and use
the regression discontinuity design. Therefore, this approach improves upon measurement, precision, and causality at the expense of generalizability.
The Zoo charges varying admission fees, so it also offers some breakdown of attendance by demographics to assess heterogeneity in responses to alerts. Separate
daily attendance is available for adults, children younger than 2, children aged
2–12, and seniors aged 62 and older, with means for each presented in Table 1. Children and the elderly are two groups considered susceptible to the effects of ozone, so
their responses may differ from adults.9 Although this definition of susceptibility is
not exhaustive, both groups are specifically targeted by air quality information. The
Zoo also offers attendance for Greater Los Angeles Zoo Association (GLAZA) members. While the Zoo is both a tourist and local attraction, GLAZA members are typically local residents who may be more aware of alerts and find it easier to switch
activities.10 Therefore, they are more likely to respond to alerts, providing one robustness check of the model.
I also collect data on attendance at baseball games (available in the ‘‘game logs’’
at www.retrosheet.org) for two major league baseball teams in SCAQMD: the Los
Angeles Dodgers, who play a short distance form downtown Los Angeles, and the
California Angels,11 who play in Anaheim, approximately 30 miles southeast of
8. The Zoo and Observatory are owned and operated by the City of Los Angeles.
9. Children’s responses to alerts could reflect state rules requiring schools to reschedule outdoor activities,
but the analysis focuses almost exclusively on summer months when children are not in school and also
includes a dummy variable for summer schedule.
10. GLAZA members pay an annual fee and do not pay an admission fee per visit.
11. The California Angels were renamed the Anaheim Angels in 1997, and then the Los Angeles Angels of
Anaheim in 2005.
Neidell
Table 1
Summary statistics
A. Daily attendance
Zoo (n ¼ 1949)
Children